Hybrid intelligent predictive maintenance model for multiclass fault classification

Data recorded from monitoring the health condition of industrial equipment are often high-dimensional, nonlinear, nonstationary and characterised by high levels uncertainty. These factors limit efficiency machine learning techniques to produce desirable results when developing effective fault classification frameworks. This paper sought propose a hybrid artificial intelligent predictive maintenance model based on Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN), Principal Component Analysis (PCA) Least Squares Support Vector Machine (LSSVM) optimised combination Coupled Simulated Annealing Nelder-Mead Simplex optimisation algorithms (ICEEMDAN-PCA-LSSVM). Here, ICEEMDAN was first employed as denoising technique decompose signals into series Intrinsic Functions (IMFs) which only relevant IMFs containing features were retained for signal reconstruction. PCA then dimension reduction through resulting set uncorrelated extracted served input LSSVM classifying various types. The proposed is compared three established methods [Linear Discriminant (LDA), (SVM) Artificial Neural Network (ANN)] multiclass capabilities. tested an experimental UCI benchmark data obtained multi-sensors hydraulic test rig. analysis revealed that ICEEMDAN-PCA-LSSVM versatile outperformed all classifiers in terms accuracy, error rate other evaluation metrics considered. drastically reduced redundancies features, allowing efficient consideration enhancement accuracy convergence speed.